

Information and codebook for replication material for "How Does Uncertainty Affect Voters' Preferences?"

***************************************************
*************** 1. FILE STRUCTURE *****************
***************************************************

1_data_preparation.R:

	prepares the exported qualtrics file ("dat_attitude_uncert.RData") for analysis. This results of the preparations are saved as "dat_uncertainty_BJPS_cleaned.RData" which is the data set used in the analysis. If you do not want to work with the raw data, then you can use dat_uncertainty_BJPS_cleaned.RData.

2_models.R 

	provides all models -- except for the multilevel modeling exercise in the appendix -- presented in the paper.

3_tables 

	contains the syntax for producing the tables in the paper. Please not that 2_models.R must be run in advance of this file.

4_graphs 

	contains the syntax for producing the graphs in the paper. Please not that 2_models.R must be run in advance of this file.

5_multilevel 

	contains the code for the multilevel models and corresponding graphs presented in the appendix.


******************************************************************************
************ 2. Codebook for "dat_uncertainty_BJPS_cleaned.RData" ************
******************************************************************************

For brevity, I mostly provide descriptions of variables whose labels are not self-explanatory. For the exact phrasing and response scales, please see the online appendix to the paper, which provides detailed information on, for example, the response scales for the belief outcome variables et cetera.

*** Explanation of suffixes ***

	There are three suffixes to the variables: _e, _p and _c. These indicate whether the respondent received expert senders (_e), partisan senders (_p) or were in the control group (_c).

*** Explanation of prefixes ***

	There are three prefixes to the variables: mw_, ct_ and _tpp. These indicate whether the questions refer to the minimum wage policy (mw), corporate tax policy (ct) or trans-pacific partnership (tpp).

*****************************************
******** Moderators and Covariates ******
*****************************************

- riskpref: the risk preferences of the individual ranging from 1 (risk averse) to 10 (risk seeking)

- partisanship: democrat (1) republican (2) independent (3)

- strong_dem: strong (1) not very strong (2)

- strong_rep: strong (1) not very strong (2)

- indep_lean: lean republican (1), lean democrat (2), neither (3)

- hhi: household income (see also the set of categorical variables generated from this)

- education: education of R (see also the set of categorical variables generated from this)

- gender: female (F) male (M)

- female: dummy indicating R is female

- age2: age squared

- politicalaffiliation: partisanship as provided by Lucid

********************************************
*********** Outcome Variables **************
********************************************

	-- For respondents in control group the variables are indicated by suffix _c. This is because the experimental design does not compare treated respondents to the control group but compare treated respondents against each other. Thus, the control group form a separate part of the data set.

	-- For respondents who were assigned to forecasts:

- *_likert: support for policy ranging from 1 (strongly oppose) to 7 (strongly favor)

- *_qual: qualitative beliefs for outcome variables 1 (increase) 2 (stay the same) 3 (decrease).

- *_qual_increase: dummy indicate belief of increasing outcome

- *_quant: quantitative beliefs for outcome variables. See supplementary material for response scale.

_ *_cert: certainty of outcome belief. lower values imply more certain about outcome belief.

- *_ideal: idealistic preferences. See supplementary material for response scale.

******************************************
********** Treatment Variables ***********
******************************************

- Condition: indicating the treatment (expert or partisan senders) or control condition of R

- *_forecast: numerical forecast presented to Rs in percentage points (MW, CT) and percentage (TPP). Note, however, that treatment is expressed in number of jobs for TPP.

- *_forecast_uncertain: dummy indicating whether respondent is exposed to two forecasts

- *_forecast_center: if one forecast: value of forecast; if two forecasts: midpoint of forecasts 

- *_forecast_uncertain_magnitude: 0 for no uncertainty, 1 for low uncertainty, 2 for high uncertainty

- *_forecast_low: numerical value of the low forecast (only for Rs who received two forecasts)

- *_forecast_high: numerical value of the high forecast (only for Rs who received two forecasts)

- *_forecast_increase: dummy indicating whether single forecast is greater than 0 (for single prediction) or lower forecast is greater than 0 (for two forecasts)

- *_forecast_decrease: dummy indicating whether single forecast is less than 0 (for single prediction) or lower forecast is less than 0 (for two forecasts)

- *_forecast_inc: dummy indicating whether forecast center is greater than 0

- *_forecast_dec: dummy indicating whether forecast center is less than 0

- *_forecast_crossing: dummy indicating whether forecast spread cross loss and gain domains

- *_forecast_spread_0: dummy indicating whether forecasts include nil prediction

- *_loss_domain: dummy indicating whether forecasts are in loss domain

- sample_period: whether in first or second sample period

- partisan: dummy variable indicating partisan senders

- tpp_treatment_partisan2: democrats send optimistic forecast (pdh) republicans send optimistic forecast (prh)

- tpp_rep_high: dummy indicating republicans sending optimistic forecast for TPP

- tpp_dem_high: dummy indicating democrats sending optimistic forecast




